Automated Multimodal Sleep Staging Using DWT-Based Wavelet Decomposition and Explainable Machine Learning with Signal Sculpting Topographies
Automatisches multimodales Sleep Staging mittels DWT-basierter Wavelet-Zerlegung und erklärbarem Machine Learning mit Signal-Sculpting-Topographien
Diagnostics · 17.08.2026 · 16 (16) · 2609
DOI: 10.3390/diagnostics16162609 ↗Auf einen Blick
In 30 gesunden Erwachsenen wurde eine automatisierte 5-Klassen-Schlafstadienklassifikation aus EEG-, EOG- und EMG-Signalen untersucht. Das beste Modell erreichte einen Macro-F1 von 0,7322 und eine Genauigkeit von 75,18 %.
Fragestellung
Wie gut kann ein automatisiertes, erklärbares multimodales Verfahren Schlafstadien aus PSG-Signalen klassifizieren?
Methode
30 gesunde Erwachsene; 1.946 annotierte 30-s-Epochen aus Sleep-EDF Expanded und Sleep Cassette; 37 Merkmale aus EEG, EOG und EMG; Vergleich von RF, SVM-RBF, GB und KNN unter stratifizierter 5-fold Cross-validation.
Ergebnisse
SVM-RBF erzielte den höchsten Macro-F1-Score von 0,7322, Cohen’s Kappa 0,6784 und eine Gesamtgenauigkeit von 75,18 %. N3 hatte den höchsten Klassen-F1 von 0,879; N1 war am schwierigsten (F1 = 0,668).
Schlussfolgerung der Autor:innen
Das vorgeschlagene Framework erreicht eine deutliche Kappa-Übereinstimmung nahe der unteren Grenze der Experten-Interrater-Reliabilität und bietet vollständige Modelltransparenz.
Kernaussagen
- SVM-RBF: Macro-F1 0,7322; Genauigkeit 75,18 %
- N3 am besten erkannt, N1 am schwierigsten
- SHAP und MDGI nannten EMG RMS, Gamma-Leistung und Permutationsentropie als Top-Merkmale
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Fachartikel in der Zeitschrift Diagnostics, online erschienen am 17.08.2026. Verfasst von Adnan Sami Sarker und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Electrical and Electronic Engineering, Chittagong University of Engineering and Technology (CUET), Chittagong 4349, Bangladesh. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Objectives
Sleep staging from polysomnographic (PSG) recordings is clinically critical for diagnosing sleep-related disorders, yet manual scoring by certified technologists remains time-consuming, costly, and subject to inter-rater variability.
Methods
This study presents an automated, explainable, and multimodal framework for five-class sleep stage classification using simultaneously acquired electroencephalography (EEG), electrooculography (EOG), and electromyography (EMG) signals. A total of 1946 annotated 30 s epochs from 30 healthy adult recording sessions (Sleep-EDF Expanded and Sleep Cassette subset) were processed through a 37-dimensional multimodal feature extraction pipeline encompassing temporal amplitude statistics, frequency-domain spectral band powers, nonlinear entropy and complexity measures, and Daubechies-4 discrete wavelet transform (DWT) energy coefficients. Four classical machine learning classifiers -Random Forest (RF), Support Vector Machine with radial basis function kernel (SVM-RBF), Gradient Boosting (GB), and K-Nearest Neighbours (KNN, k = 7) were benchmarked under stratified five-fold cross-validation.
Results
SVM-RBF achieved the highest macro-averaged F1-score of 0.7322 (Cohen’s kappa 0.6784, overall accuracy 75.18%). N3 deep slow-wave sleep achieved the highest per-class F1 of 0.879, while N1 light sleep was the most challenging (F1 = 0.668). SHapley Additive exPlanations (SHAP) and RF mean decrease in Gini impurity (MDGI) analysis jointly identified EMG root mean square amplitude (MDGI = 0.0805), gamma band power (0.0784), and permutation entropy (0.0434) as the three most discriminative features. As a novel methodological contribution, sixteen categories of signal sculpting visualisations were developed, translating abstract multivariate features into clinically interpretable graphical representations.
Conclusions
The proposed framework achieves substantial kappa agreement approaching the lower bound of expert inter-rater reliability (0.76–0.82) while providing full model transparency, with direct implications for wearable sleep monitoring device design.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Electrical and Electronic Engineering, Chittagong University of Engineering and Technology (CUET), Chittagong 4349, Bangladesh.
- University Dental College Dhaka, Dhaka 1212, Bangladesh.
- Khan Dental Clinic, Dhaka 1212, Bangladesh.
- ELITE Research Lab, Queens, NY 11435, USA.
- Centre for Intelligent Cloud Computing (CICC), COE of Advanced Cloud, Faculty of Information Science and Technology, Multimedia University, Lama Bukit Beruang, Melaka 75450, Malaysia.
ORCID-Profile: Tze Hui Liew
Zitieren
APA
Sarker, A. S., Samir, K. M. M., Shakib, Z. K., Morol, M. K., & Liew, T. H. (2026). Automated Multimodal Sleep Staging Using DWT-Based Wavelet Decomposition and Explainable Machine Learning with Signal Sculpting Topographies. Diagnostics, 16(16), 2609. https://doi.org/10.3390/diagnostics16162609
Vancouver
Sarker AS, Samir KMM, Shakib ZK, Morol MK, Liew TH. Automated Multimodal Sleep Staging Using DWT-Based Wavelet Decomposition and Explainable Machine Learning with Signal Sculpting Topographies. Diagnostics. 2026;16(16):2609. doi:10.3390/diagnostics16162609
BibTeX
@article{Sarker2026,
author = {Sarker, Adnan Sami and Samir, Kazi Mahatir Mohammed and Shakib, Zunayed Khan and Morol, Md Kishor and Liew, Tze Hui},
title = {Automated Multimodal Sleep Staging Using DWT-Based Wavelet Decomposition and Explainable Machine Learning with Signal Sculpting Topographies},
journal = {Diagnostics},
year = {2026},
volume = {16},
number = {16},
pages = {2609},
doi = {10.3390/diagnostics16162609}
}
EUVIMED